Embed Size (px)
Citation preview
Optimizing the Canadian Health Workforce
Ottawa, Ontario
October 4, 2016
Jean Moore, DrPH, MSN
Center for Health Workforce Studies
School of Public Health | University at Albany, SUNY
Workforce Supply/Demand Forecast Modeling in the
US: Can Microsimulation Help Us Break Out of Our
Siloes?
The Power of Projection Models........
2
0
2,000
4,000
6,000
8,000
10,000
12,000
19
96
19
97
19
98
19
99
20
00
20
01
20
02
20
03
20
04
20
05
20
06
20
07
20
08
20
09
20
10
20
11
20
12
20
13
20
14
20
15
Associates
Bachelor's
Total
New York RN Graduations, by Degree Type, 1996-2015
Source: Center for Health Workforce Studies
Historical Background on the Federally Supported
Workforce Supply/Demand Models
3
• Siloed models (separate models for different occupations)
• Different contractors built different models using different platforms, methods and
assumptions
• Static models—parameters constant over time and across states
• Separate supply and demand models
• Infrequently updated
• Limited capability to analyze policy or emerging care delivery models
• Limited ability to capture geographic variation in population risk factors
Nursing Supply Model ● Nursing Demand Model ● Physician Supply Model ● Integrated Requirements Model ● Pharmacist Supply and Requirements Model ● Dental
Requirements Model ● General Services Demand Model ● other misc. models
Health Workforce Simulation Model
Before Now
Health Workforce Simulation Model:
Design Criteria
• Built on solid theoretical underpinnings
• Dynamic model that can integrate professions and link supply
with demand
• Can account for both current and future availability of data
• Can be adapted for analysis at state or local levels
• Easy to maintain/update as new data become available
• Supports scenario modeling
4
Microsimulation Approach for Modeling Workforce
Demand
• Individual patients are the unit of observation
o Predict use of health care services by individual
o Determine how care will be provided to individuals
o Sum across individuals to produce aggregate statistics
• Approach
o Develop population health database with health profile for representative
sample of the population
o Develop predictive equations (using regression analysis) to model health care
use
• Translate health care encounters into demand for practitioners
– Use data on how practitioners divide their time between care delivery settings and
patient encounters to create estimates of patient encounters per full time
equivalent
5
Health Profile for Each Person in
Stratified Random Sample
6
Demographics & Socioeconomics
• Demographics
o Age
o Sex
o Race/ethnicity
• Socioeconomics
o Household income
o Insurance (private, public non-Medicare,
Medicare, uninsured)
Risk Factors & Chronic Conditions
• Obese/overweight*
• Smoking status *
• Diagnosed with o Hypertension *
o High cholesterol *
o Coronary heart disease *
o Diabetes *
o History of stroke *
o History of cancer *
o Asthma
o Arthritis *
Key Data Sources
• Center for Disease Control and Prevention: Behavioral Risk Factor Surveillance System (2011-2013 data);
NY EpiQuery
• Census Bureau: American Community Survey and population projections (2013)
• Medical Expenditure Panel Survey and National Inpatient Sample (2013)
* Information available for adults only
Example: Use of
Cardiology Services
1 Rate ratios from Poisson
regression analysis using 2009-
2013 MEPS/2013 NIS.
2 Odds ratios from logistic
regression analysis using 2009-
2013 MEPS.
Statistically significant at the
0.05 (*) or 0.01 (**) level.
Health Risk & Behavior
Economic & Policy
Care Delivery
Demographics
Cardiologist Cardiology-related Primary Diagnosis
Parameter
Office
Visits1
Outpatient
Visits1
Emergency
Visits2
Hospital-
ization2
Inpatient
Days1
Rac
e-
Eth
nic
ity Non-Hispanic White 1.00 1.00 1.00 1.00 1.00
Non-Hispanic Black 0.79** 0.97 1.36** 1.32** 1.14**
Non-Hispanic Other 0.90** 0.75** 0.86 0.94 1.10**
Hispanic 0.79** 0.68** 0.93 0.84** 1.07**
Male 1.13** 1.59** 0.89* 1.11 0.97**
Age
18-34 years 0.11** 0.24** 0.66** 0.40** 0.84**
35-44 years 0.22** 0.63** 0.95 0.76** 0.80**
45-64 years 0.50** 0.86** 1.05 1.10 0.86**
65-74 years 0.83** 1.21** 1.11 1.50** 0.93**
75+ years 1.00** 1.00** 1.00** 1.00** 1.00
Smoker 0.73** 0.84** 1.22** 1.11
Dia
gno
sed
wit
h
Hypertension 1.55** 1.13** 3.86** 2.66**
Heart disease 8.50** 10.73** 2.93** 3.84**
History of heart attack 1.63** 1.36** 2.36** 2.60**
History of stroke 1.08** 1.26** 2.92** 3.04**
Diabetes 1.15** 1.34** 1.01 1.19** 1.02**
Arthritis 1.10** 1.24** 0.96 0.96
Asthma 1.04* 1.08** 1.00 1.07
History of cancer 1.06** 1.11** 1.01 0.99
Bo
dy
Wei
ght
Normal 1.00** 1.00** 1.00** 1.00**
Overweight 1.04** 1.09** 0.87** 0.82**
Obese 1.11** 1.18** 1.01 1.02
Insu
red
Has insurance 2.61** 2.09** 0.92 1.09 0.99*
In Medicaid 1.36** 1.30** 1.59** 1.71** 1.23**
In managed care plan 1.00 1.24** 0.99 0.99
Ho
use
ho
ld In
com
e
<$10,000 0.90** 0.97 1.23** 1.19**
$10,000 to <$15,000 0.92** 0.91** 1.16* 1.20**
$15,000 to < $20,000 0.93** 0.93* 0.82 0.99
$20,000 to < $25,000 0.89** 0.73** 1.15 1.06
$25,000 to < $35,000 0.92** 0.96 1.16* 1.05
$35,000 to < $50,000 0.88** 1.07* 0.91 0.93
$50,000 to < $75,000 0.96* 1.17** 0.93 0.82**
$75,000 or higher 1.00 1.00 1.00 1.00
Metro Area 1.31** 1.09** 1.07 0.91 1.03**
Care Delivery Patterns: Converting Service Demand to
Health Profession FTEs
• 1,000 ambulatory visits to a pediatrician equates to approximately 0.23 FTE
pediatrician; 1,000 hospital rounds equates to approximately 0.48 FTE
pediatrician
• Every 4,469 visits to a physician’s office translates to 1 full time
equivalent RN
8
Microsimulation Approach to
Workforce Supply Modeling
9
• Individuals are the unit of observation
• Modeling process
o Starts with database containing starting year workforce supply
o Each year to 2030, model:
– New entrants to the workforce
– Workforce attrition (retirement, mortality, out migration)
– Other activities (labor force participation, hours worked, geographic mobility by
occupation/specialty and provider demographics)
o End of year supply = starting supply for subsequent year
• Influencing factors
o Demographics of the workforce
o Economic and policy factors (e.g., earnings, payment system)
Nursing Workforce Simulation Model:
Supply Component
• Simulate likely career choices of individual clinicians
o Microsimulation—modeling workforce decisions of individual clinicians, rather
than stock-and-flow models that simulate groups of clinicians
• Dynamic modeling
o Environmental and market factors—clinicians respond to changes in the
economy, healthcare operating environment, and policy
o Shortages/surpluses affect clinician workforce decisions
• Workforce activities: what, where, how, when
o What type of work will I do?
o Where will I work (e.g., state of practice)?
o How many hours will I work?
o When will I retire?
10
Conceptual Model for Nurse Workforce Supply
11
Scenario Modeling Capability
• What if….
o Supply declines? (fewer new grads, early retirements)
o Supply increases? (more new grads, delayed retirements)
• What if….
o Demand changes
– Increase in the number of people with health insurance
– Improved chronic disease management
– Used new technology supports better access to services (e.g., telehealth)
– Reduced the number of unnecessary emergency room visits or hospitalizations
• Can model a wide range of scenarios—reflecting uncertainties in future trends in both supply and demand
12
Can This Be Used to Model Team Based Care?
• Can estimate demand across professions with similar clinical roles
and responsibilities
o physicians, nurse practitioners, physician assistants
o dentists, dental therapists
• Can’t easily track which team member provided which clinical service
to a patient
• Can’t account for non-clinical services provided by non-licensed
workers (i.e., community health workers, care coordinators) that
provide non-clinical services
13
Limitations
• Lack of data
o Supply data of any kind on most professions challenging to find
o Detailed demand data to better understand impacts of team based models of
care
• Lack of consistency in membership on team based care delivery
models
• National and state level assessments fail to account for local
supply/demand imbalances
• Doesn’t account for state-to-state scope of practice variation
• No consensus on the right benchmark to use
14
Looking Ahead
• Continued federal funding over the next four years to use the
microsimulation model to forecast workforce supply/demand
imbalances in:
o Long-term care
o Allied health
o Oral health
o Primary care
15
